Jaehyeok Kim

I'm a Ph.D. student at Hong Kong University of Science and Technology (HKUST) advised by Prof. Dan Xu, with Hong Kong PhD Fellowship. I received my Bachelor's degree in Computer Science also from HKUST with Academic Achievement Medal (top 1%, 49 recipients). I was fortunate to work as a research intern at Naver Clova.

My research interests lie in computer vision, 3D scene reconstruction and editing, and dynamic human modeling with 3D representations like NeRF or 3DGS.

Email  /  Scholar  /  X (Twitter)  /  Github

profile photo

News!

  • 01 May, 2026: One paper accepted at ICML 2026! See you in Seoul 🇰🇷
  • 26 June, 2025: One paper accepted at ICCV 2025!
  • 01 July, 2024: One paper accepted at ECCV 2024. Appreciate Prof. Dan Xu's great help.
  • Publication

    C-DIC overview
    Context-Driven Incremental Compression for Multi-Turn Dialogue Generation
    Yeongseo Jung, Jaehyeok Kim, Eunseo Jung, Jiachuan Wang, Yongqi Zhang, Ka Chun Cheung, Simon See, Lei Chen
    ICML 2026
    arXiv / Code (Coming soon!)

    C-DIC introduces revisable thread-wise dialogue memory with an incremental retrieve-revise-write loop to improve long-horizon dialogue quality while keeping latency stable over increasing number of dialogue turns.

    From One to More: Contextual Part Latents for 3D Generation
    Shaocong Dong*, Lihe Ding*, Xiao Chen, Yaokun Li, Yuxin Wang, Yucheng Wang, Qi Wang, Jaehyeok Kim, Chenjian Gao, Zhanpeng Huang, Zibin Wang, Tianfan Xue†, Dan Xu†
    ICCV 2025
    project page / arXiv / Code

    CoPart is a part-based 3D generation framework that uses contextual part latents and a mutual guidance strategy to enable coherent, controllable 3D object creation with fine-grained part-level conditioning.

    MoCo-NeRF: Motion-Oriented Compositional Neural Radiance Fields for Monocular Dynamic Human Modeling
    Jaehyeok Kim, Dongyoon Wee, Dan Xu†
    ECCV 2024
    project page / arXiv

    Modeling non-rigid motions of dynamic humans as radiance residuals reduces learning complexity and enhances rendering quality, allowing for efficient and accurate reconstruction of detailed cloth dynamics.

    Activities

  • Conference Reviewer: ICCV, BMVC, ACM MM
  • Teaching

  • Teaching Assistant: Deep Perception, Localization, and Planning for Autonomous Vehicles @ HKUST in Spring 2023
  • Teaching Assistant: Deep 2D and 3D Visual Scene Understanding @ HKUST in Spring 2024 & Spring 2026

  • Source code from Jon Barron.